Speech-to-Speech Question-Answering on WebQ
61.5AccuracyDIFFUSPEECH
Evaluation results
| Method | Links | |
|---|---|---|
| DIFFUSPEECHType=Diff., S→S capability=true2026.01 | 61.5 | |
| Qwen2.5-Omni-7BModel Parameters=7B2026.03 | 60.1 | |
| VocalNetModel Parameters=8B2026.03 | 59.8 | |
| MinMoType=AR, S→S capability=true2026.01 | 55 | |
| VIRBABackbone=Qwen2.5-Omni2025.08 | 51.3 | |
| LLaMA-Omni2Model Parameters=7B2026.03 | 49.2 | |
| VIRBABackbone=Qwen2-Audio2025.08 | 47.9 | |
| MoshiType=AR, S→S capability=true2026.01 | 47.5 | |
| GLM-4-VoiceModel Parameters=9B2026.03 | 46.9 | |
| SPEECH-OMNI-LITEModel Parameters=8B2026.03 | 46.2 | |
| Step-Audio-R12025.08 | 44.5 | |
| Kimi-Audio2025.08 | 43.2 | |
| AF3Full name=Audio Flamingo 32025.08 | 42.8 | |
| Single-view RL2025.08 | 40.7 | |
| Llama-Omni2Type=AR, S→S capability=true2026.01 | 40.4 | |
| Qwen2.5-Omni2025.08 | 38.6 | |
| ELLSAmode=S2S2025.10 | 36.5 | |
| Lychee-FDType=Native2026.07 | 33.9 | |
| Baichuan-Omni-1.5Model Parameters=7B2026.03 | 32.39 | |
| GLM-4-Voice2025.08 | 32.2 | |
| Lychee-FD (w/o Sem-Channel)Type=Native2026.07 | 31.5 | |
| StepAudio-2-miniType=Half-duplex2026.07 | 30.8 | |
| Freeze-OmniModel Parameters=7B2026.03 | 29.8 | |
| VITA 1.5Type=System-level2026.07 | 29.2 | |
| SALMONN-omniType=Native2026.07 | 28.1 | |
| Freeze-Omnimode=S2S2025.10 | 27.9 | |
| Freeze-OmniType=System-level2026.07 | 25.8 | |
| Fun-Audio-ChatType=Native2026.07 | 24.4 | |
| Lychee-FD (w/o Param-Sep)Type=Native2026.07 | 22.5 | |
| Moshimode=S2S2025.10 | 22.1 | |
| MoshiType=Native2026.07 | 19.6 | |
| FLM-audioType=Native2026.07 | 14.5 | |
| SpiritLMType=AR, S→S capability=true2026.01 | 12.3 | |
| Qwen2-Audio base2025.08 | 10.1 | |
| SpeechGPTType=AR, S→S capability=true2026.01 | 6.3 | |
| dGSLMType=Native2026.07 | 0.2 |